azure-openai-provider

Vercel AI Provider for running Large Language Models using Azure OpenAI

Note: This module currently supports chat models and text inputs. Completion and embeddings is not currently implemented.

Note: This module is under development and may contain errors and frequent incompatible changes.

All releases will be of type MAJOR following the 0.MAJOR.MINOR scheme. Only bugs and non-breaking updates will be released as MINOR. Please read the Tested models and capabilities section to know about the features implemented in this provider.

Installation

The Azure OpenAI provider is available in the azure-openai-provider module. You can install it with

npm i azure-openai-provider
yarn add azure-openai-provider

Provider Instance

Unlike other providers, there is no default instance and you must create your own instance:

import { createAzureOpenAI } from 'azure-openai-provider';

const azureOpenAI = createAzureOpenAI({
  resourceName: 'YourAzureOpenAIResourceName',
  // or baseURL if using a proxy. baseURL: 'https://apigateway.acme.com'
});

await streamText({
  model: azureOpenAi('modelId', 'yourDeploymentName'),
  ...
})

You can use the following optional settings to customize the Azure OpenAI provider instance:

• resourceName string

  Required if baseURL is not specified. The provider will format the base url to be https://${resourceName}.azure.openai.com

• baseURL string

  Use a different URL prefix for API calls, e.g. to use proxy servers. If baseURL is passed it will override the resourceName property. Note: the provider will append /openai/deployments/${this.deploymentName} to the end of the baseURL ex. baseURL: "https://apigateway.acme.com"

• headers Record<string,string>

  Custom headers to include in the requests.

About

This package is a slightly modified version of the official OpenAI provider by Vercel.

Summary of current changes: Added apiVersion and resourceName to the provider options Mapped apiKey to the api-key header, instead of the 'Authorization' header Change convert-to-openai-chat-message.ts to map user messages as {role: "user", content: string} instead of the new format. This will likely break vision capable models

Models

The first argument is you the openai model id. This is not currently utilized but could be in the future.

The second argument is your azure deployment name, e.g. gpt-4-turbo-deployment.

const model = azureOpenAI('gpt-4-1106-preview', 'gpt-4-turbo-deployment');

Examples

Inside the examples folder, you will find some example projects to see how the provider works. Each folder has its own README with the usage description.

Tested capabilities

AI SDK Core

✅ - confirmed working ❌ - does not work

| Method | Status | | -------------- | -------------------------- | | generateText | ✅ | | streamText | ✅ | | generateObject | ✅ | | streamObject | untested (assumed working) |

AI SDK UI

| Method | Status | | ------------- | ------------------------------ | | useChat | ✅ | | useCompletion | ❌ | | useAssistant | untested (assumed not working) |

AI SDK RSC

| Method | Status | | ------------------ | ------ | | createStreamableUI | ✅ |